Recording intelligenceAI-generated brief · check the source for context

Leopold Aschenbrenner OMR Hedge Fund Collapse

1:30:27 recording · EN · 2 speakers

Listen to the original

Listen to the episode
Executive Summary AI
  • The episode picks up the OMR podcast discussion of Leopold Aschenbrenner, the Situational Awareness author who left school at 15, graduated top of his class at Columbia at 19, researched with Nobel laureate Josef Stieglitz, worked at FTX's charity arm and then OpenAI before starting his own hedge fund.1:52
  • His fund started with 225 million from LPs including the Collison twins, Daniel Gross and Ned Friedman, ran an early long bet on the AI supply chain at roughly four- to fivefold leverage, and was reported at up to 45 billion under management within about two years.3:10
  • Then margin calls forced him to liquidate every public-market position at once after names such as Sandys fell almost 45%, Nebius about minus 30%, CoreWeave minus 25% and Core Scientific minus 15, with the hosts noting a 20% drop on five-times leverage wipes out all the collateral.5:09
  • Ken Griffin's Citadel, with 71 billion under management, bought the book out in a block trade at a 5 to 10 percent discount and saw those same stocks jump 8 to 12 percent the next day, while investors are probably still in the plus after a 450 percent year.11:31
  • On the model side, Sam Altman is announcing a price offensive with the 5.6 family, where Luna is said to be 80% cheaper and Terra 20% cheaper, and the earnings round shows Microsoft adding around 500 billion in market value in a day, Meta down over 10%, and AWS accelerating to 37% growth at a 42 billion quarter with a 40% operating margin.20:48

Brief overview

A 450% AI-rally hedge fund collapsed on leverage, while cloud earnings and falling model prices boomed.

  1. Leverage, not insight, ended the runA four- to fivefold lever turned 20-30% drawdowns in chip and energy names into margin calls that liquidated the whole public book.
  2. Liquidity itself is the product Citadel sellsCitadel took the portfolio at a 5 to 10 percent discount and the stocks rose 8 to 12 percent the next day once the selling pressure was gone.
  3. AI is the first software that gets cheaper fastLuna at 80% cheaper and Terra at 20% cheaper fit the pattern that last year's performance costs a tenth today, unlike SaaS pricing which never really fell.

Questions this recording answers

10 questions, each answered where it is said
Why did Leopold Aschenbrenner's hedge fund have to sell all its public positions?

The fund was over-leveraged, roughly a factor of four to five, and got margin calls. Positions fell sharply from the end of Q1: Sandys almost 45%, Bloom Energy and Nebius about 30%, CoreWeave 25%, Core Scientific 15%. At that leverage a 20% loss wipes out 100% of the collateral.

Answered around 5:21
How big did Situational Awareness grow, and who backed it?

It started with 225 million, with founding LPs including the Collison twins of Stripe, Daniel Gross and Nat Friedman. Within about two years it reportedly reached around 25 billion, with one report of 45 billion under management, built by betting early on the AI supply chain: chips, energy and turbines.

Answered around 3:10
What role did Citadel play in the fund's liquidation?

Ken Griffin's Citadel reportedly offered to buy the whole portfolio in a block trade at a 5 to 10 percent discount, so it would not be dumped on the market. The hosts call it free money for being liquid, and note the stocks then jumped 8 to 12 percent the same day.

Answered around 12:49
Why did Aschenbrenner leave OpenAI?

He was said to have leaked information. He had warned internally that OpenAI's security was not good enough for such powerful technology, and shared thoughts and documents with outside researchers. He argues the state must control AI like the atomic bomb. It is unclear whether his views or the sharing caused his exit.

Answered around 18:03
How much cheaper are OpenAI's new Luna and Terra models?

OpenAI announced a price offensive in its 5.6 family, named Luna, Terra and Sol from smallest to largest. Luna is said to be 80% cheaper, almost down to open-weights level, and Terra 20% cheaper, attributed to internal efficiency gains. The hosts see about half of research now going into making models cheaper.

Answered around 20:33
Did OpenAI really make as much in July as in all of Q2?

The hosts doubt it. July was a very good month, but tripling the average Q2 month would be surprising. They think it is far more likely that the newly added annualized revenue in July exceeded that of the whole of Q2, which was weak for new ARR. OpenAI still grows roughly threefold a year.

Answered around 35:51
What drove Microsoft's big jump after earnings?

Microsoft gained about 15.5%, around 500 billion dollars in one day. Service revenue grew 23% to 90 billion, operating margin reached 45% and operating cash flow 55 billion. Cloud growth accelerated from 40 to 43%, slightly above expectations, and remaining performance obligations grew 84%.

Answered around 45:38
Why did Meta's stock fall despite 28% revenue growth?

Growth slowed from 33% to 28%, operating margin dropped partly due to a 2.4 to 2.5 billion court payment, and free cash flow fell 90% to close to a billion after about 30 billion in capex. The market dislikes Meta's top-tier AI spending without a clear strategy for using it.

Answered around 47:49
How fast is AWS growing now?

AWS accelerated from 20% to 24% to 28% over three quarters and now 37%, well above the 31 to 32% analysts expected. It makes 42 billion in quarterly sales at about a 40% operating margin, 17 billion in operating income, and Amazon raised its capex outlook, mainly because storage and chips cost more.

Answered around 1:01:18
Do Robinhood users beat the market?

No. For the third quarter in a row the average Robinhood user underperformed, by roughly 2.4 to 2.5% by the host's rough calculation. One dollar invested like Robinhood users would now be 1.76 dollars, versus 3.60 in the Nasdaq and 2.40 in the S&P 500.

Answered around 1:13:02
Key Quote

“As I said, a 20% loss is enough to erase 100% of your money.”

— Philipp Klöckner, Leopold Aschenbrenner OMR Hedge Fund Collapse, at 7:07▶ listen at 7:07
Key Quote

“What doesn't work is to saw a AI strategy in such an old, sclerotic, crumbled organization mediator.”

— Philipp Klöckner, Leopold Aschenbrenner OMR Hedge Fund Collapse, at 28:35▶ listen at 28:35
Key Quote

“So 10 billion a month is still spent on AI.”

— Philipp Klöckner, Leopold Aschenbrenner OMR Hedge Fund Collapse, at 29:49▶ listen at 29:49
Key Quote

“The smartest thing Entropic could do would be to invent a button that says, for $2, this request would be in within three seconds.”

— Philipp Klöckner, Leopold Aschenbrenner OMR Hedge Fund Collapse, at 1:06:16▶ listen at 1:06:16

Cite this

“Leopold Aschenbrenner OMR Hedge Fund Collapse.” https://dcs-cached.megaphone.fm/LINDALA5816227906.mp3. Transcript summary by WhipScribe, https://whipscribe.com/transcript-pages/leopold-aschenbrenner-omr-hedge-fund-collapse.html.

Report this page
What is wrong with this page?